There’s a lot of outdated information on the Web that leads new PHP users astray, propagating bad practices and insecure code. PHP: The Right Way is an easy-to-read, quick reference for PHP popular coding standards, links to authoritative tutorials around the Web and what the contributors consider to be best practices at the present time.
There is no canonical way to use PHP. This website aims to introduce new PHP developers to some topics which they may not discover until it is too late, and aims to give seasoned pros some fresh ideas on those topics they’ve been doing for years without ever reconsidering. This website will also not tell you which tools to use, but instead offer suggestions for multiple options, when possible explaining the differences in approach and use-case.
This is a living document and will continue to be updated with more helpful information and examples as they become available.
PHP: The Right Way is translated into many different languages:
The most recent version of PHP: The Right Way is also available in PDF, EPUB and MOBI formats. Go to Leanpub
Help make this website the best resource for new PHP programmers! Contribute on GitHub
If you are getting started with PHP, start with the current stable release of PHP 7.2. PHP 7.x adds many new features over the older 5.x versions. The engine has been largely re-written, and PHP is now even quicker than older versions.
You should try to upgrade to the latest stable version quickly - PHP 5.6 will not receive security updates beyond 2018. Upgrading is easy, as there are not many backwards compatibility breaks. If you are not sure which version a function or feature is in, you can check the PHP documentation on the php.net website.
With PHP 5.4 or newer, you can start learning PHP without installing and configuring a full-fledged web server. To start the server, run the following command from your terminal in your project’s web root:
macOS comes prepackaged with PHP but it is normally a little behind the latest stable release. There are multiple ways to install the latest PHP version on macOS.
Homebrew is a package manager for macOS that helps you easily install PHP and various extensions. The Homebrew core repository provides “formulae” for PHP 5.6, 7.0, 7.1, and 7.2. Install the latest version with this command:
brew install email@example.com
You can switch between Homebrew PHP versions by modifying your
PATH variable. Alternatively, you can use brew-php-switcher to switch PHP versions automatically.
The MacPorts Project is an open-source community initiative to design an easy-to-use system for compiling, installing, and upgrading either command-line, X11 or Aqua based open-source software on the OS X operating system.
MacPorts supports pre-compiled binaries, so you don’t need to recompile every dependency from the source tarball files, it saves your life if you don’t have any package installed on your system.
At this point, you can install
php71 using the
port install command, for example:
sudo port install php56 sudo port install php71
And you can run
select command to switch your active PHP:
sudo port select --set php php71
phpbrew is a tool for installing and managing multiple PHP versions. This can be really useful if two different applications/projects require different versions of PHP, and you are not using virtual machines.
Another popular option is php-osx.liip.ch which provides one liner installation methods for versions 5.3 through 7.3. It doesn’t overwrite the PHP binaries installed by Apple, but installs everything in a separate location (/usr/local/php5).
Another option that gives you control over the version of PHP you install, is to compile it yourself. In that case be sure to have installed either Xcode or Apple’s substitute “Command Line Tools for XCode” downloadable from Apple’s Mac Developer Center.
The solutions listed above mainly handle PHP itself, and do not supply things like Apache, Nginx or a SQL server. “All-in-one” solutions such as MAMP and XAMPP will install these other bits of software for you and tie them all together, but ease of setup comes with a trade-off of flexibility.
You can download the binaries from windows.php.net/download. After the extraction of PHP, it is recommended to set the PATH to the root of your PHP folder (where php.exe is located) so you can execute PHP from anywhere.
For learning and local development, you can use the built in webserver with PHP 5.4+ so you don’t need to worry about configuring it. If you would like an “all-in-one” which includes a full-blown webserver and MySQL too then tools such as the Web Platform Installer, XAMPP, EasyPHP, OpenServer and WAMP will help get a Windows development environment up and running fast. That said, these tools will be a little different from production so be careful of environment differences if you are working on Windows and deploying to Linux.
If you need to run your production system on Windows, then IIS7 will give you the most stable and best performance. You can use phpmanager (a GUI plugin for IIS7) to make configuring and managing PHP simple. IIS7 comes with FastCGI built in and ready to go, you just need to configure PHP as a handler. For support and additional resources there is a dedicated area on iis.net for PHP.
Generally running your application on different environment in development and production can lead to strange bugs popping up when you go live. If you are developing on Windows and deploying to Linux (or anything non-Windows) then you should consider using a Virtual Machine.
Chris Tankersley has a very helpful blog post on what tools he uses to do PHP development using Windows.
A common question among those starting out with writing programs for the web is, “where do I put my stuff?” Over the years, this answer has consistently been “where the
DocumentRoot is.” Although this answer is not complete, it’s a great place to start.
For security reasons, configuration files should not be accessible by a site’s visitors; therefore, public scripts are kept in a public directory and private configurations and data are kept outside of that directory.
For each team, CMS, or framework one works in, a standard directory structure is used by each of those entities. However, if one is starting a project alone, knowing which filesystem structure to use can be daunting.
Paul M. Jones has done some fantastic research into common practices of tens of thousands of github projects in the realm of PHP. He has compiled a standard file and directory structure, the Standard PHP Package Skeleton, based on this research. In this directory structure,
DocumentRoot should point to
public/, unit tests should be in the
tests/ directory, and third party libraries, as installed by composer, belong in the
vendor/ directory. For other files and directories, abiding by the Standard PHP Package Skeleton will make the most sense to contributors of a project.
The PHP community is large and diverse, composed of innumerable libraries, frameworks, and components. It is common for PHP developers to choose several of these and combine them into a single project. It is important that PHP code adhere (as close as possible) to a common code style to make it easy for developers to mix and match various libraries for their projects.
The Framework Interop Group has proposed and approved a series of style recommendations. Not all of them related to code-style, but those that do are PSR-1, PSR-2 and PSR-4. These recommendations are merely a set of rules that many projects like Drupal, Zend, Symfony, Laravel, CakePHP, phpBB, AWS SDK, FuelPHP, Lithium, etc are adopting. You can use them for your own projects, or continue to use your own personal style.
Ideally, you should write PHP code that adheres to a known standard. This could be any combination of PSRs, or one of the coding standards made by PEAR or Zend. This means other developers can easily read and work with your code, and applications that implement the components can have consistency even when working with lots of third-party code.
You can fix the code layout automatically by using one of the following tools:
And you can run phpcs manually from shell:
phpcs -sw --standard=PSR2 file.php
It will show errors and describe how to fix them. It can also be helpful to include this command in a git hook. That way, branches which contain violations against the chosen standard cannot enter the repository until those violations have been fixed.
If you have PHP_CodeSniffer, then you can fix the code layout problems reported by it, automatically, with the PHP Code Beautifier and Fixer.
phpcbf -w --standard=PSR2 file.php
Another option is to use the PHP Coding Standards Fixer. It will show which kind of errors the code structure had before it fixed them.
php-cs-fixer fix -v --level=psr2 file.php
English is preferred for all symbol names and code infrastructure. Comments may be written in any language easily readable by all current and future parties who may be working on the codebase.
Finally, a good supplementary resource for writing clean PHP code is Clean Code PHP.
PHP is a flexible, dynamic language that supports a variety of programming techniques. It has evolved dramatically over the years, notably adding a solid object-oriented model in PHP 5.0 (2004), anonymous functions and namespaces in PHP 5.3 (2009), and traits in PHP 5.4 (2012).
PHP has a very complete set of object-oriented programming features including support for classes, abstract classes, interfaces, inheritance, constructors, cloning, exceptions, and more.
PHP supports first-class functions, meaning that a function can be assigned to a variable. Both user-defined and built-in functions can be referenced by a variable and invoked dynamically. Functions can be passed as arguments to other functions (a feature called Higher-order Functions) and functions can return other functions.
Recursion, a feature that allows a function to call itself, is supported by the language, but most PHP code is focused on iteration.
New anonymous functions (with support for closures) are present since PHP 5.3 (2009).
PHP 5.4 added the ability to bind closures to an object’s scope and also improved support for callables such that they can be used interchangeably with anonymous functions in almost all cases.
PHP supports various forms of meta-programming through mechanisms like the Reflection API and Magic Methods. There are
many Magic Methods available like
__invoke(), etc. that allow
developers to hook into class behavior. Ruby developers often say that PHP is lacking
method_missing, but it is
As mentioned above, the PHP community has a lot of developers creating lots of code. This means that one library’s PHP code might use the same class name as another. When both libraries are used in the same namespace, they collide and cause trouble.
Namespaces solve this problem. As described in the PHP reference manual, namespaces may be compared to operating system directories that namespace files; two files with the same name may co-exist in separate directories. Likewise, two PHP classes with the same name may co-exist in separate PHP namespaces. It’s as simple as that.
It is important for you to namespace your code so that it may be used by other developers without fear of colliding with other libraries.
One recommended way to use namespaces is outlined in PSR-4, which aims to provide a standard file, class and namespace convention to allow plug-and-play code.
In October 2014 the PHP-FIG deprecated the previous autoloading standard: PSR-0. Both PSR-0 and PSR-4 are still perfectly usable. The latter requires PHP 5.3, so many PHP 5.2-only projects implement PSR-0.
If you’re going to use an autoloader standard for a new application or package, look into PSR-4.
The Standard PHP Library (SPL) is packaged with PHP and provides a collection of classes and interfaces. It is made up primarily of commonly needed datastructure classes (stack, queue, heap, and so on), and iterators which can traverse over these datastructures or your own classes which implement SPL interfaces.
PHP was created to write web applications, but is also useful for scripting command line interface (CLI) programs. Command line PHP programs can help automate common tasks like testing, deployment, and application administration.
CLI PHP programs are powerful because you can use your app’s code directly without having to create and secure a web GUI for it. Just be sure not to put your CLI PHP scripts in your public web root!
Try running PHP from your command line:
-i option will print your PHP configuration just like the
-a option provides an interactive shell, similar to ruby’s IRB or python’s interactive shell. There are a number
of other useful command line options, too.
Let’s write a simple “Hello, $name” CLI program. To try it out, create a file named
hello.php, as below.
PHP sets up two special variables based on the arguments your script is run with.
$argc is an integer
variable containing the argument count and
$argv is an array variable containing each argument’s value.
The first argument is always the name of your PHP script file, in this case
exit() expression is used with a non-zero number to let the shell know that the command failed. Commonly used
exit codes can be found here.
To run our script, above, from the command line:
One of the most useful tools in software development is a proper debugger. It allows you to trace the execution of your code and monitor the contents of the stack. Xdebug, PHP’s debugger, can be utilized by various IDEs to provide Breakpoints and stack inspection. It can also allow tools like PHPUnit and KCacheGrind to perform code coverage analysis and code profiling.
If you find yourself in a bind, willing to resort to
print_r(), and you still can’t find the solution -
maybe you need to use the debugger.
Installing Xdebug can be tricky, but one of its most important features is “Remote Debugging” - if you develop code locally and then test it inside a VM or on another server, Remote Debugging is the feature that you will want to enable right away.
Traditionally, you will modify your Apache VHost or .htaccess file with these values:
The “remote host” and “remote port” will correspond to your local computer and the port that you configure your IDE to listen on. Then it’s just a matter of putting your IDE into “listen for connections” mode, and loading the URL:
Your IDE will now intercept the current state as the script executes, allowing you to set breakpoints and probe the values in memory.
Graphical debuggers make it very easy to step through code, inspect variables, and eval code against the live runtime. Many IDE’s have built-in or plugin-based support for graphical debugging with Xdebug. MacGDBp is a free, open-source, stand-alone Xdebug GUI for Mac.
There are a ton of PHP libraries, frameworks, and components to choose from. Your project will likely use several of them — these are project dependencies. Until recently, PHP did not have a good way to manage these project dependencies. Even if you managed them manually, you still had to worry about autoloaders. That is no longer an issue.
Currently there are two major package management systems for PHP - Composer and PEAR. Composer is currently the most popular package manager for PHP, however for a long time PEAR was the primary package manager in use. Knowing PEAR’s history is a good idea, since you may still find references to it even if you never use it.
Composer is the recommended dependency manager for PHP. List your project’s dependencies in a
composer.json file and,
with a few simple commands, Composer will automatically download your project’s dependencies and setup autoloading for
you. Composer is analogous to NPM in the node.js world, or Bundler in the Ruby world.
There is a plethora of PHP libraries that are compatible with Composer and ready to be used in your project. These “packages” are listed on Packagist, the official repository for Composer-compatible PHP libraries.
The safest way to download composer is by following the official instructions.
This will verify the installer is not corrupt or tampered with.
The installer installs a
composer.phar binary in your current working directory.
We recommend installing Composer globally (e.g. a single copy in
/usr/local/bin). To do so, run this command next:
Note: If the above fails due to permissions, prefix with
To run a locally installed Composer you’d use
php composer.phar, globally it’s simply
For Windows users the easiest way to get up and running is to use the ComposerSetup installer, which
performs a global install and sets up your
$PATH so that you can just call
composer from any
directory in your command line.
Composer keeps track of your project’s dependencies in a file called
composer.json. You can manage it
by hand if you like, or use Composer itself. The
composer require command adds a project dependency
and if you don’t have a
composer.json file, one will be created. Here’s an example that adds Twig
as a dependency of your project.
composer init command will guide you through creating a full
for your project. Either way, once you’ve created your
composer.json file you can tell Composer to
download and install your dependencies into the
vendor/ directory. This also applies to projects
you’ve downloaded that already provide a
Next, add this line to your application’s primary PHP file; this will tell PHP to use Composer’s autoloader for your project dependencies.
Now you can use your project dependencies, and they’ll be autoloaded on demand.
Composer creates a file called
composer.lock which stores the exact version of each package it
downloaded when you first ran
composer install. If you share your project with others,
composer.lock file is included, so that when they run
composer install they’ll
get the same versions as you. To update your dependencies, run
composer update. Don’t use
composer update when deploying, only
composer install, otherwise you may end up with different
package versions on production.
This is most useful when you define your version requirements flexibly. For instance, a version
~1.8 means “anything newer than
1.8.0, but less than
2.0.x-dev”. You can also use
* wildcard as in
1.8.*. Now Composer’s
composer update command will upgrade all your
dependencies to the newest version that fits the restrictions you define.
To receive notifications about new version releases you can sign up for libraries.io, a web service that can monitor dependencies and send you alerts on updates.
The Security Advisories Checker is a web service and a command-line tool, both will examine your
file and tell you if you need to update any of your dependencies.
Composer can also handle global dependencies and their binaries. Usage is straight-forward, all you need
to do is prefix your command with
global. If for example you wanted to install PHPUnit and have it
available globally, you’d run the following command:
This will create a
~/.composer folder where your global dependencies reside. To have the installed
packages’ binaries available everywhere, you’d then add the
~/.composer/vendor/bin folder to your
A veteran package manager that some PHP developers enjoy is PEAR. It behaves similarly to Composer, but has some notable differences.
PEAR requires each package to have a specific structure, which means that the author of the package must prepare it for usage with PEAR. Using a project which was not prepared to work with PEAR is not possible.
PEAR installs packages globally, which means after installing them once they are available to all projects on that server. This can be good if many projects rely on the same package with the same version but might lead to problems if version conflicts between two projects arise.
You can install PEAR by downloading the
.phar installer and executing it. The PEAR documentation has
detailed install instructions for every operating system.
If you are using Linux, you can also have a look at your distribution package manager. Debian and Ubuntu,
for example, have an apt
If the package is listed on the PEAR packages list, you can install it by specifying the official name:
If the package is hosted on another channel, you need to
discover the channel first and also specify it when
installing. See the Using channel docs for more information on this topic.
If you are already using Composer and you would like to install some PEAR code too, you can use Composer to
handle your PEAR dependencies. This example will install code from
The first section
"repositories" will be used to let Composer know it should “initialize” (or “discover” in PEAR
terminology) the pear repo. Then the
require section will prefix the package name like this:
The “pear” prefix is hardcoded to avoid any conflicts, as a pear channel could be the same as another packages vendor name for example, then the channel short name (or full URL) can be used to reference which channel the package is in.
When this code is installed it will be available in your vendor directory and automatically available through the Composer autoloader:
To use this PEAR package simply reference it like so:
PHP is a vast language that allows coders of all levels the ability to produce code not only quickly, but efficiently. However, while advancing through the language, we often forget the basics that we first learnt (or overlooked) in favor of short cuts and/or bad habits. To help combat this common issue, this section is aimed at reminding coders of the basic coding practices within PHP.
PHP has a class named DateTime to help you when reading, writing, comparing or calculating with date and time. There are many date and time related functions in PHP besides DateTime, but it provides nice object-oriented interface to most common uses. It can handle time zones, but that is outside this short introduction.
To start working with DateTime, convert raw date and time string to an object with
createFromFormat() factory method
new DateTime to get the current date and time. Use
format() method to convert DateTime back to a string for
Calculating with DateTime is possible with the DateInterval class. DateTime has methods like
take a DateInterval as an argument. Do not write code that expect same number of seconds in every day, both daylight
saving and timezone alterations will break that assumption. Use date intervals instead. To calculate date difference
diff() method. It will return new DateInterval, which is super easy to display.
On DateTime objects you can use standard comparison:
One last example to demonstrate the DatePeriod class. It is used to iterate over recurring events. It can take two DateTime objects, start and end, and the interval for which it will return all events in between.
A popular PHP API extension is Carbon. It inherits everything in the DateTime class, so involves minimal code alterations, but extra features include Localization support, further ways to add, subtract and format a DateTime object, plus a means to test your code by simulating a date and time of your choosing.
When you are building your application it is helpful to use common patterns in your code and common patterns for the overall structure of your project. Using common patterns is helpful because it makes it much easier to manage your code and lets other developers quickly understand how everything fits together.
If you use a framework then most of the higher level code and project structure will be based on that framework, so a lot of the pattern decisions are made for you. But it is still up to you to pick out the best patterns to follow in the code you build on top of the framework. If, on the other hand, you are not using a framework to build your application then you have to find the patterns that best suit the type and size of application that you’re building.
You can learn more about PHP design patterns and see working examples at:
Right now PHP does not support Unicode at a low level. There are ways to ensure that UTF-8 strings are processed OK, but it’s not easy, and it requires digging in to almost all levels of the web app, from HTML to SQL to PHP. We’ll aim for a brief, practical summary.
The basic string operations, like concatenating two strings and assigning strings to variables, don’t need anything
special for UTF-8. However, most string functions, like
strlen(), do need special consideration. These
functions often have an
mb_* counterpart: for example,
mb_* strings are made
available to you via the Multibyte String Extension, and are specifically designed to operate on Unicode strings.
You must use the
mb_* functions whenever you operate on a Unicode string. For example, if you use
substr() on a
UTF-8 string, there’s a good chance the result will include some garbled half-characters. The correct function to use
would be the multibyte counterpart,
The hard part is remembering to use the
mb_* functions at all times. If you forget even just once, your Unicode
string has a chance of being garbled during further processing.
Not all string functions have an
mb_* counterpart. If there isn’t one for what you want to do, then you might be out
You should use the
mb_internal_encoding() function at the top of every PHP script you write (or at the top of your
global include script), and the
mb_http_output() function right after it if your script is outputting to a browser.
Explicitly defining the encoding of your strings in every script will save you a lot of headaches down the road.
Additionally, many PHP functions that operate on strings have an optional parameter letting you specify the character
encoding. You should always explicitly indicate UTF-8 when given the option. For example,
htmlentities() has an
option for character encoding, and you should always specify UTF-8 if dealing with such strings. Note that as of PHP 5.4.0, UTF-8 is the default encoding for
Finally, If you are building a distributed application and cannot be certain that the
mbstring extension will be
enabled, then consider using the patchwork/utf8 Composer package. This will use
mbstring if it is available, and
fall back to non UTF-8 functions if not.
If your PHP script accesses MySQL, there’s a chance your strings could be stored as non-UTF-8 strings in the database even if you follow all of the precautions above.
To make sure your strings go from PHP to MySQL as UTF-8, make sure your database and tables are all set to the
utf8mb4 character set and collation, and that you use the
utf8mb4 character set in the PDO connection string. See
example code below. This is critically important.
Note that you must use the
utf8mb4 character set for complete UTF-8 support, not the
utf8 character set! See
Further Reading for why.
mb_http_output() function to ensure that your PHP script outputs UTF-8 strings to your browser.
The browser will then need to be told by the HTTP response that this page should be considered as UTF-8. Today, it is common to set the character set in the HTTP response header like this:
The historic approach to doing that was to include the charset
<meta> tag in your page’s
Disclaimer for newcomers: i18n and l10n are numeronyms, a kind of abbreviation where numbers are used to shorten words - in our case, internationalization becomes i18n and localization, l10n.
First of all, we need to define those two similar concepts and other related things:
The easiest way to internationalize PHP software is by using array files and using those strings in templates, such as
<h1><?=$TRANS['title_about_page']?></h1>. This way is, however, hardly recommended for serious projects, as it poses
some maintenance issues along the road - some might appear in the very beginning, such as pluralization. So, please,
don’t try this if your project will contain more than a couple of pages.
The most classic way and often taken as reference for i18n and l10n is a Unix tool called
gettext. It dates
back to 1995 and is still a complete implementation for translating software. It is easy enough to get running, while
still sporting powerful supporting tools. It is about Gettext we will be talking here. Also, to help you not get messy
over the command-line, we will be presenting a great GUI application that can be used to easily update your l10n source
There are common libraries used that support Gettext and other implementations of i18n. Some of them may seem easier to install or sport additional features or i18n file formats. In this document, we focus on the tools provided with the PHP core, but here we list others for completion:
intlextension (including pluralized messages).
gettextcommand), and can also export to other formats besides
Other frameworks also include i18n modules, but those are not available outside of their codebases:
@langhelper for template files.
Intlextension, available since PHP 5.3, and based on the ICU project; this enables Yii to run powerful replacements, like spelling out numbers, formatting dates, times, intervals, currency, and ordinals.
If you decide to go for one of the libraries that provide no extractors, you may want to use the gettext formats, so you can use the original gettext toolchain (including Poedit) as described in the rest of the chapter.
You might need to install Gettext and the related PHP library by using your package manager, like
After installed, enable it by adding
extension=gettext.so (Linux/Unix) or
extension=php_gettext.dll (Windows) to
Here we will also be using Poedit to create translation files. You will probably find it in your system’s package manager; it is available for Unix, Mac, and Windows, and can be downloaded for free on their website as well.
There are three files you usually deal with while working with gettext. The main ones are PO (Portable Object) and MO (Machine Object) files, the first being a list of readable “translated objects” and the second, the corresponding binary to be interpreted by gettext when doing localization. There’s also a POT (Template) file, which simply contains all existing keys from your source files, and can be used as a guide to generate and update all PO files. Those template files are not mandatory: depending on the tool you are using to do l10n, you can go just fine with only PO/MO files. You will always have one pair of PO/MO files per language and region, but only one POT per domain.
There are some cases, in big projects, where you might need to separate translations when the same words convey different meaning given a context. In those cases, you split them into different domains. They are, basically, named groups of POT/PO/MO files, where the filename is the said translation domain. Small and medium-sized projects usually, for simplicity, use only one domain; its name is arbitrary, but we will be using “main” for our code samples. In Symfony projects, for example, domains are used to separate the translation for validation messages.
A locale is simply a code that identifies one version of a language. It is defined following the ISO 639-1 and ISO 3166-1 alpha-2 specs: two lower-case letters for the language, optionally followed by an underline and two upper-case letters identifying the country or regional code. For rare languages, three letters are used.
For some speakers, the country part may seem redundant. In fact, some languages have dialects in different
countries, such as Austrian German (
de_AT) or Brazilian Portuguese (
pt_BR). The second part is used to distinguish
between those dialects - when it is not present, it is taken as a “generic” or “hybrid” version of the language.
To use Gettext, we will need to adhere to a specific structure of folders. First, you will need to select an arbitrary
root for your l10n files in your source repository. Inside it, you will have a folder for each needed locale, and a
LC_MESSAGES folder that will contain all your PO/MO pairs. Example:
As we said in the introduction, different languages might sport different plural rules. However, gettext saves us from
this trouble once again. When creating a new
.po file, you will have to declare the plural rules for that
language, and translated pieces that are plural-sensitive will have a different form for each of those rules. When
calling Gettext in code, you will have to specify the number related to the sentence, and it will work out the correct
form to use - even using string substitution if needed.
Plural rules include the number of plurals available and a boolean test with
n that would define in which rule the
given number falls (starting the count with 0). For example:
nplurals=1; plural=0- only one rule
nplurals=2; plural=(n != 1);- two rules, first if N is one, second rule otherwise
nplurals=2; plural=(n > 1);- two rules, second if N is bigger than one, first otherwise
Now that you understood the basis of how plural rules works - and if you didn’t, please look at a deeper explanation on the LingoHub tutorial -, you might want to copy the ones you need from a list instead of writing them by hand.
When calling out Gettext to do localization on sentences with counters, you will have to give him the
related number as well. Gettext will work out what rule should be in effect and use the correct localized version.
You will need to include in the
.po file a different sentence for each plural rule defined.
After all that theory, let’s get a little practical. Here’s an excerpt of a
.po file - don’t mind with its format,
but with the overall content instead; you will learn how to edit it easily later:
The first section works like a header, having the
msgstr especially empty. It describes the file encoding,
plural forms and other things that are less relevant.
The second section translates a simple string from English to
Brazilian Portuguese, and the third does the same, but leveraging string replacement from
sprintf so the
translation may contain the user name and visit date.
The last section is a sample of pluralization forms, displaying
the singular and plural version as
msgid in English and their corresponding translations as
msgstr 0 and 1
(following the number given by the plural rule). There, string replacement is used as well so the number can be seen
directly in the sentence, by using
%d. The plural forms always have two
msgid (singular and plural), so it is
advised not to use a complex language as the source of translation.
As you might have noticed, we are using as source ID the actual sentence in English. That
msgid is the same used
throughout all your
.po files, meaning other languages will have the same format and the same
msgid fields but
Talking about translation keys, there are two main “schools” here:
msgidas a real sentence. The main advantages are:
msgidacross several language files.
msgidas a unique, structured key. It would describe the sentence role in the application in a structured way, including the template or part where the string is located instead of its content.
en.pofile, that translators would read to understand what to write in
Hello there, User!on the said untranslated French page). That is good it as would force translation to be complete before publishing - however, bad as translation issues would be remarkably awful in the interface. Some libraries, though, include an option to specify a given language as “fallback”, having a similar behavior as the other approach.
The Gettext manual favors the first approach as, in general, it is easier for translators and users in case of trouble. That is how we will be working here as well. However, the Symfony documentation favors keyword-based translation, to allow for independent changes of all translations without affecting templates as well.
In a typical application, you would use some Gettext functions while writing static text in your pages. Those sentences
would then appear in
.po files, get translated, compiled into
.mo files and then, used by Gettext when rendering
the actual interface. Given that, let’s tie together what we have discussed so far in a step-by-step example:
gettext()simply translates a
msgidinto its corresponding
msgstrfor a given language. There’s also the shorthand function
_()that works the same way;
ngettext()does the same but with plural rules;
dngettext(), that allows you to override the domain for a single call. More on domain configuration in the next example.
i18n_setup.phpas used above), selecting the correct locale and configuring Gettext
One of the great advantages Gettext has over custom framework i18n packages is its extensive and powerful file format. “Oh man, that’s quite hard to understand and edit by hand, a simple array would be easier!” Make no mistake, applications like Poedit are here to help - a lot. You can get the program from their website, it’s free and available for all platforms. It’s a pretty easy tool to get used to, and a very powerful one at the same time - using all features Gettext has available. This guide is based on PoEdit 1.8.
In the first run, you should select “File > New…” from the menu. You’ll be asked straight ahead for the language:
here you can select/filter the language you want to translate to, or use that format we mentioned before, such as
Now, save the file - using that directory structure we mentioned as well. Then you should click “Extract from sources”, and here you’ll configure various settings for the extraction and translation tasks. You’ll be able to find all those later through “Catalog > Properties”:
gettext()(and siblings) are called - this is usually your templates/views folder(s). This is the only mandatory setting;
gettext()and similar function calls look like in several programming languages, but you might as well create your own translation functions. It will be here you’ll add those other methods. This will be discussed later in the “Tips” section.
After setting those points it will run a scan through your source files to find all the localization calls. After every scan PoEdit will display a summary of what was found and what was removed from the source files. New entries will fed empty into the translation table, and you’ll start typing in the localized versions of those strings. Save it and a .mo file will be (re)compiled into the same folder and ta-dah: your project is internationalized.
As you may have noticed before, there are two main types of localized strings: simple ones and those with plural forms. The first ones have simply two boxes: source and localized string. The source string cannot be modified as Gettext/Poedit do not include the powers to alter your source files - you should change the source itself and rescan the files. Tip: you may right-click a translation line and it will hint you with the source files and lines where that string is being used. On the other hand, plural form strings include two boxes to show the two source strings, and tabs so you can configure the different final forms.
Whenever you change your sources and need to update the translations, just hit Refresh and Poedit will rescan the code, removing non-existent entries, merging the ones that changed and adding new ones. It may also try to guess some translations, based on other ones you did. Those guesses and the changed entries will receive a “Fuzzy” marker, indicating it needs review, appearing golden in the list. It is also useful if you have a translation team and someone tries to write something they are not sure about: just mark Fuzzy, and someone else will review later.
Finally, it is advised to leave “View > Untranslated entries first” marked, as it will help you a lot to not forget any entry. From that menu, you can also open parts of the UI that allow you to leave contextual information for translators if needed.
If you are running PHP as a module on Apache (
mod_php), you might face issues with the
.mo file being cached. It
happens the first time it is read, and then, to update it, you might need to restart the server. On Nginx and PHP5 it
usually takes only a couple of page refreshes to refresh the translation cache, and on PHP7 it is rarely needed.
As preferred by many people, it is easier to use
_() instead of
gettext(). Many custom i18n libraries from
frameworks use something similar to
t() as well, to make translated code shorter. However, that is the only function
that sports a shortcut. You might want to add in your project some others, such as
or maybe a fancy
_r() that would join
sprintf() calls. Other libraries, such as
oscarotero’s Gettext also provide helper functions like these.
In those cases, you’ll need to instruct the Gettext utility on how to extract the strings from those new functions.
Don’t be afraid; it is very easy. It is just a field in the
.po file, or a Settings screen on Poedit. In the editor,
that option is inside “Catalog > Properties > Source keywords”. Remember: Gettext already knows the default functions
for many languages, so don’t be afraid if that list seems empty. You need to include there the specifications of those
new functions, following a specific format:
t()that simply returns the translation for a string, you can specify it as
t. Gettext will know the only function argument is the string to be translated;
__('one user', '%d users', $number), the specification would be
__:1,2, meaning the first form is the first argument, and the second form is the second argument. If your number comes as the first argument instead, the spec would be
__:2,3, indicating the first form is the second argument, and so on.
After including those new rules in the
.po file, a new scan will bring in your new strings just as easy as before.
Dependency injection is a software design pattern that allows the removal of hard-coded dependencies and makes it possible to change them, whether at run-time or compile-time.
This quote makes the concept sound much more complicated than it actually is. Dependency Injection is providing a component with its dependencies either through constructor injection, method calls or the setting of properties. It is that simple.
We can demonstrate the concept with a simple, yet naive example.
Here we have a
Database class that requires an adapter to speak to the database. We instantiate the adapter in the
constructor and create a hard dependency. This makes testing difficult and means the
Database class is very tightly
coupled to the adapter.
This code can be refactored to use Dependency Injection and therefore loosen the dependency.
Now we are giving the
Database class its dependency rather than creating it itself. We could even create a method
that would accept an argument of the dependency and set it that way, or if the
$adapter property was
could set it directly.
If you have ever read about Dependency Injection then you have probably seen the terms “Inversion of Control” or “Dependency Inversion Principle”. These are the complex problems that Dependency Injection solves.
Inversion of Control is as it says, “inverting the control” of a system by keeping organizational control entirely separate from our objects. In terms of Dependency Injection, this means loosening our dependencies by controlling and instantiating them elsewhere in the system.
For years, PHP frameworks have been achieving Inversion of Control, however, the question became, which part of control are we inverting, and where to? For example, MVC frameworks would generally provide a super object or base controller that other controllers must extend to gain access to its dependencies. This is Inversion of Control, however, instead of loosening dependencies, this method simply moved them.
Dependency Injection allows us to more elegantly solve this problem by only injecting the dependencies we need, when we need them, without the need for any hard coded dependencies at all.
The Single Responsibility Principle is about actors and high-level architecture. It states that “A class should have only one reason to change.” This means that every class should only have responsibility over a single part of the functionality provided by the software. The largest benefit of this approach is that it enables improved code reusability. By designing our class to do just one thing, we can use (or re-use) it in any other program without changing it.
The Open/Closed Principle is about class design and feature extensions. It states that “Software entities (classes, modules, functions, etc.) should be open for extension, but closed for modification.” This means that we should design our modules, classes and functions in a way that when a new functionality is needed, we should not modify our existing code but rather write new code that will be used by existing code. Practically speaking, this means that we should write classes that implement and adhere to interfaces, then type-hint against those interfaces instead of specific classes.
The largest benefit of this approach is that we can very easily extend our code with support for something new without having to modify existing code, meaning that we can reduce QA time, and the risk for negative impact to the application is substantially reduced. We can deploy new code, faster, and with more confidence.
The Liskov Substitution Principle is about subtyping and inheritance. It states that “Child classes should never break the parent class’ type definitions.” Or, in Robert C. Martin’s words, “Subtypes must be substitutable for their base types.”
For example, if we have a
FileInterface interface which defines an
embed() method, and we have
classes which both implement the
embed() method, then we can expect that the usage of the
embed() method will always
do the thing that we intend. If we later create a
Gist class which implement the
interface, we will already know and understand what the
embed() method will do. The largest benefit of this approach
is that we have the ability to build flexible and easily-configurable programs, because when we change one object of a
FileInterface) to another we don’t need to change anything else in our program.
The Interface Segregation Principle (ISP) is about business-logic-to-clients communication. It states that “No client should be forced to depend on methods it does not use.” This means that instead of having a single monolithic interface that all conforming classes need to implement, we should instead provide a set of smaller, concept-specific interfaces that a conforming class implements one or more of.
For example, a
Bus class would be interested in a
steeringWheel() method, but a
class would not. Conversely, a
Tricycle class would be interested in a
handlebars() method, but a
Bus class would not. There is no need to have all of these types of vehicles implement support for both
steeringWheel() as well as
handlebars(), so we should break-apart the source interface.
The Dependency Inversion Principle is about removing hard-links between discrete classes so that new functionality can be leveraged by passing a different class. It states that one should “Depend on Abstractions. Do not depend on concretions.”. Put simply, this means our dependencies should be interfaces/contracts or abstract classes rather than concrete implementations. We can easily refactor the above example to follow this principle.
There are several benefits to the
Database class now depending on an interface rather than a concretion.
Consider that we are working in a team and the adapter is being worked on by a colleague. In our first example, we would have to wait for said colleague to finish the adapter before we could properly mock it for our unit tests. Now that the dependency is an interface/contract we can happily mock that interface knowing that our colleague will build the adapter based on that contract.
An even bigger benefit to this method is that our code is now much more scalable. If a year down the line we decide that we want to migrate to a different type of database, we can write an adapter that implements the original interface and injects that instead, no more refactoring would be required as we can ensure that the adapter follows the contract set by the interface.
The first thing you should understand about Dependency Injection Containers is that they are not the same thing as Dependency Injection. A container is a convenience utility that helps us implement Dependency Injection, however, they can be and often are misused to implement an anti-pattern, Service Location. Injecting a DI container as a Service Locator in to your classes arguably creates a harder dependency on the container than the dependency you are replacing. It also makes your code much less transparent and ultimately harder to test.
Most modern frameworks have their own Dependency Injection Container that allows you to wire your dependencies together through configuration. What this means in practice is that you can write application code that is as clean and de- coupled as the framework it is built on.
Many times your PHP code will use a database to persist information. You have a few options to connect and interact with your database. The recommended option until PHP 5.1.0 was to use native drivers such as mysqli, pgsql, mssql, etc.
Native drivers are great if you are only using one database in your application, but if, for example, you are using MySQL and a little bit of MSSQL, or you need to connect to an Oracle database, then you will not be able to use the same drivers. You’ll need to learn a brand new API for each database — and that can get silly.
The mysql extension for PHP is incredibly old and has been superseded by two other extensions:
To save digging into your
php.ini settings to see which module you are using, one option is to search for
in your editor of choice. If any functions such as
mysql_query() show up, then
Even if you are not using PHP 7.x yet, failing to consider this upgrade as soon as possible will lead to greater hardship when the PHP 7.x upgrade does come about. The best option is to replace mysql usage with mysqli or PDO in your applications within your own development schedules so you won’t be rushed later on.
If you are upgrading from mysql to mysqli, beware lazy upgrade guides that suggest you can simply find and replace
mysqli_*. Not only is that a gross oversimplification, it misses out on the advantages that mysqli provides, such as parameter binding, which is also offered in PDO.
PDO is a database connection abstraction library — built into PHP since 5.1.0 — that provides a common interface to talk with many different databases. For example, you can use basically identical code to interface with MySQL or SQLite:
PDO will not translate your SQL queries or emulate missing features; it is purely for connecting to multiple types of database with the same API.
PDO allows you to safely inject foreign input (e.g. IDs) into your SQL queries without worrying
about database SQL injection attacks.
This is possible using PDO statements and bound parameters.
Let’s assume a PHP script receives a numeric ID as a query parameter. This ID should be used to fetch a user record
from a database. This is the
wrong way to do this:
This is terrible code. You are inserting a raw query parameter into a SQL query. This will get you hacked in a
heartbeat, using a practice called SQL Injection. Just imagine if a hacker passes in an inventive
id parameter by
calling a URL like
http://domain.com/?id=1%3BDELETE+FROM+users. This will set the
$_GET['id'] variable to
FROM users which will delete all of your users! Instead, you should sanitize the ID input using PDO bound parameters.
This is correct code. It uses a bound parameter on a PDO statement. This escapes the foreign input ID before it is introduced to the database preventing potential SQL injection attacks.
You should also be aware that database connections use up resources and it was not unheard-of to have resources exhausted if connections were not implicitly closed, however this was more common in other languages. Using PDO you can implicitly close the connection by destroying the object by ensuring all remaining references to it are deleted, i.e. set to NULL. If you don’t do this explicitly, PHP will automatically close the connection when your script ends - unless of course you are using persistent connections.
When developers first start to learn PHP, they often end up mixing their database interaction up with their presentation logic, using code that might look like this:
This is bad practice for all sorts of reasons, mainly that it’s hard to debug, hard to test, hard to read and it is going to output a lot of fields if you don’t put a limit on there.
Consider the most basic step:
That is a good start. Put those two items in two different files and you’ve got some clean separation.
Create a class to place that method in and you have a “Model”. Create a simple
.php file to put the presentation
logic in and you have a “View”, which is very nearly MVC - a common OOP architecture for most
This is essentially the same as what most modern frameworks are doing, albeit a little more manual. You might not need to do all of that every time, but mixing together too much presentation logic and database interaction can be a real problem if you ever want to unit-test your application.
Many frameworks provide their own abstraction layer which may or may not sit on top of PDO. These will often emulate features for one database system that is missing from another by wrapping your queries in PHP methods, giving you actual database abstraction instead of just the connection abstraction that PDO provides. This will of course add a little overhead, but if you are building a portable application that needs to work with MySQL, PostgreSQL and SQLite then a little overhead will be worth it the sake of code cleanliness.
Templates provide a convenient way of separating your controller and domain logic from your presentation logic. Templates typically contain the HTML of your application, but may also be used for other formats, such as XML. Templates are often referred to as “views”, which make up part of the second component of the model–view–controller (MVC) software architecture pattern.
The main benefit to using templates is the clear separation they create between the presentation logic and the rest of your application. Templates have the sole responsibility of displaying formatted content. They are not responsible for data lookup, persistence or other more complex tasks. This leads to cleaner, more readable code which is especially helpful in a team environment where developers work on the server-side code (controllers, models) and designers work on the client-side code (markup).
Templates also improve the organization of presentation code. Templates are typically placed in a “views” folder, each defined within a single file. This approach encourages code reuse where larger blocks of code are broken into smaller, reusable pieces, often called partials. For example, your site header and footer can each be defined as templates, which are then included before and after each page template.
Finally, depending on the library you use, templates can offer more security by automatically escaping user-generated content. Some libraries even offer sand-boxing, where template designers are only given access to white-listed variables and functions.
Plain PHP templates are simply templates that use native PHP code. They are a natural choice since PHP is actually a template language itself. That simply means that you can combine PHP code within other code, like HTML. This is beneficial to PHP developers as there is no new syntax to learn, they know the functions available to them, and their code editors already have PHP syntax highlighting and auto-completion built-in. Further, plain PHP templates tend to be very fast as no compiling stage is required.
Every modern PHP framework employs some kind of template system, most of which use plain PHP by default. Outside of frameworks, libraries like Plates or Aura.View make working with plain PHP templates easier by offering modern template functionality such as inheritance, layouts and extensions.
Using the Plates library.
Using the Plates library.
While PHP has evolved into a mature, object oriented language, it hasn’t improved much as a templating language. Compiled templates, like Twig, Brainy, or Smarty*, fill this void by offering a new syntax that has been geared specifically to templating. From automatic escaping, to inheritance and simplified control structures, compiled templates are designed to be easier to write, cleaner to read and safer to use. Compiled templates can even be shared across different languages, Mustache being a good example of this. Since these templates must be compiled there is a slight performance hit, however this is very minimal when proper caching is used.
*While Smarty offers automatic escaping, this feature is NOT enabled by default.
Using the Twig library.
Using the Twig library.
In many “exception-heavy” programming languages, whenever anything goes wrong an exception will be thrown. This is certainly a viable way to do things, but PHP is an “exception-light” programming language. While it does have exceptions and more of the core is starting to use them when working with objects, most of PHP itself will try to keep processing regardless of what happens, unless a fatal error occurs.
This is only a notice error, and PHP will happily carry on. This can be confusing for those coming from “exception-heavy” languages, because referencing a missing variable in Python for example will throw an exception:
The only real difference is that Python will freak out over any small thing, so that developers can be super sure any potential issue or edge-case is caught, whereas PHP will keep on processing unless something extreme happens, at which point it will throw an error and report it.
PHP has several levels of error severity. The three most common types of messages are errors, notices and warnings.
These have different levels of severity;
E_WARNING. Errors are fatal run-time errors and
are usually caused by faults in your code and need to be fixed as they’ll cause PHP to stop executing. Notices are
advisory messages caused by code that may or may not cause problems during the execution of the script, execution is
not halted. Warnings are non-fatal errors, execution of the script will not be halted.
Another type of error message reported at compile time are
E_STRICT messages. These messages are used to suggest
changes to your code to help ensure best interoperability and forward compatibility with upcoming versions of PHP.
Error Reporting can be changed by using PHP settings and/or PHP function calls. Using the built in PHP function
error_reporting() you can set the level of errors for the duration of the script execution by passing one of the
predefined error level constants, meaning if you only want to see Errors and Warnings - but not Notices - then you can
You can also control whether or not errors are displayed to the screen (good for development) or hidden, and logged (good for production). For more information on this check out the Error Reporting section.
You can also tell PHP to suppress specific errors with the Error Control Operator
@. You put this operator at the
beginning of an expression, and any error that’s a direct result of the expression is silenced.
This will output
$foo['bar'] if it exists, but will simply return a null and print nothing if the variable
'bar' key does not exist. Without the error control operator, this expression could create a
PHP Notice: Undefined
variable: foo or
PHP Notice: Undefined index: bar error.
This might seem like a good idea, but there are a few undesirable tradeoffs. PHP handles expressions using an
@ in a
less performant way than expressions without an
@. Premature optimization may be the root of all programming
arguments, but if performance is particularly important for your application/library it’s important to understand the
error control operator’s performance implications.
Secondly, the error control operator completely swallows the error. The error is not displayed, and the error is not sent to the error log. Also, stock/production PHP systems have no way to turn off the error control operator. While you may be correct that the error you’re seeing is harmless, a different, less harmless error will be just as silent.
If there’s a way to avoid the error suppression operator, you should consider it. For example, our code above could be rewritten like this:
One instance where error suppression might make sense is where
fopen() fails to find a file to load. You could check
for the existence of the file before you try to load it, but if the file is deleted after the check and before the
fopen() (which might sound impossible, but it can happen) then
fopen() will return false and throw an error. This
is potentially something PHP should resolve, but is one case where error suppression might seem like the only valid
Earlier we mentioned there’s no way in a stock PHP system to turn off the error control operator. However, Xdebug has
xdebug.scream ini setting which will disable the error control operator. You can set this via your
with the following.
You can also set this value at runtime with the
The “Scream” PHP extension offers similar functionality to Xdebug’s, although Scream’s ini setting is named
This is most useful when you’re debugging code and suspect an informative error is suppressed. Use scream with care, and as a temporary debugging tool. There’s lots of PHP library code that may not work with the error control operator disabled.
PHP is perfectly capable of being an “exception-heavy” programming language, and only requires a few lines of code to
make the switch. Basically you can throw your “errors” as “exceptions” using the
ErrorException class, which extends
This is a common practice implemented by a large number of modern frameworks such as Symfony and Laravel. In debug mode (or dev mode) both of these frameworks will display a nice and clean stack trace.
There are also some packages available for better error and exception handling and reporting. Like Whoops!, which comes with the default installation of Laravel and can be used in any framework as well.
By throwing errors as exceptions in development you can handle them better than the usual result, and if you see an exception during development you can wrap it in a catch statement with specific instructions on how to handle the situation. Each exception you catch instantly makes your application that little bit more robust.
More information on this and details on how to use
ErrorException with error handling can be found at
Exceptions are a standard part of most popular programming languages, but they are often overlooked by PHP programmers. Languages like Ruby are extremely Exception heavy, so whenever something goes wrong such as a HTTP request failing, or a DB query goes wrong, or even if an image asset could not be found, Ruby (or the gems being used) will throw an exception to the screen meaning you instantly know there is a mistake.
PHP itself is fairly lax with this, and a call to
file_get_contents() will usually just get you a
FALSE and a
Many older PHP frameworks like CodeIgniter will just return a false, log a message to their proprietary logs and maybe
let you use a method like
$this->upload->get_error() to see what went wrong. The problem here is that you have to go
looking for a mistake and check the docs to see what the error method is for this class, instead of having it made
Another problem is when classes automatically throw an error to the screen and exit the process. When you do this you stop another developer from being able to dynamically handle that error. Exceptions should be thrown to make a developer aware of an error; they then can choose how to handle this. E.g.:
Exception class provides very little debugging context for the developer; however, to remedy this, it is
possible to create a specialized
Exception type by sub-classing the generic
This means you can add multiple catch blocks and handle different Exceptions differently. This can lead to the creation of a lot of custom Exceptions, some of which could have been avoided using the SPL Exceptions provided in the SPL extension.
If for example you use the
__call() Magic Method and an invalid method is requested then instead of throwing a
standard Exception which is vague, or creating a custom Exception just for that, you could just
throw new BadMethodCallException;.
It is very important for every PHP developer to learn the basics of web application security, which can be broken down into a handful of broad topics:
There are bad people ready and willing to exploit your web application. It is important that you take necessary precautions to harden your web application’s security. Luckily, the fine folks at The Open Web Application Security Project (OWASP) have compiled a comprehensive list of known security issues and methods to protect yourself against them. This is a must read for the security-conscious developer. Survive The Deep End: PHP Security by Padraic Brady is also another good web application security guide for PHP.
Eventually everyone builds a PHP application that relies on user login. Usernames and passwords are stored in a database and later used to authenticate users upon login.
Hashing is an irreversible, one-way function. This produces a fixed-length string that cannot be feasibly reversed. This means you can compare a hash against another to determine if they both came from the same source string, but you cannot determine the original string. If passwords are not hashed and your database is accessed by an unauthorized third-party, all user accounts are now compromised.
Unlike hashing, encryption is reversible (provided you have the key). Encryption is useful in other areas, but is a poor strategy for securely storing passwords.
Passwords should also be individually salted by adding a random string to each password before hashing. This prevents dictionary attacks and the use of “rainbow tables” (a reverse list of crytographic hashes for common passwords.)
Hashing and salting are vital as often users use the same password for multiple services and password quality can be poor.
Additionally, you should use a specialized password hashing algoithm rather than fast, general-purpose cryptographic hash function (e.g. SHA256). The short list of acceptable password hashing algorithms (as of June 2018) to use are:
Fortunately, nowadays PHP makes this easy.
Hashing passwords with
In PHP 5.5
password_hash() was introduced. At this time it is using BCrypt, the strongest algorithm currently
supported by PHP. It will be updated in the future to support more algorithms as needed though. The
library was created to provide forward compatibility for PHP >= 5.3.7.
Below we hash a string, and then check the hash against a new string. Because our two source strings are different (‘secret-password’ vs. ‘bad-password’) this login will fail.
password_hash() takes care of password salting for you. The salt is stored, along with the algorithm and “cost”, as part of the hash.
password_verify() extracts this to determine how to check the password, so you don’t need a separate database field to store your salts.
Never ever (ever) trust foreign input introduced to your PHP code. Always sanitize and validate foreign input before
using it in code. The
filter_input() functions can sanitize text and validate text formats (e.g.
Foreign input can be anything:
$_POST form input data, some values in the
$_SERVER superglobal, and the
HTTP request body via
fopen('php://input', 'r'). Remember, foreign input is not limited to form data submitted by the
user. Uploaded and downloaded files, session values, cookie data, and data from third-party web services are foreign
While foreign data can be stored, combined, and accessed later, it is still foreign input. Every time you process, output, concatenate, or include data in your code, ask yourself if the data is filtered properly and can it be trusted.
Data may be filtered differently based on its purpose. For example, when unfiltered foreign input is passed into HTML
very dangerous attack. One way to avoid XSS is to sanitize all user-generated data before outputting it to your page by
removing HTML tags with the
strip_tags() function or escaping characters with special meaning into their respective
HTML entities with the
Another example is passing options to be executed on the command line. This can be extremely dangerous (and is usually
a bad idea), but you can use the built-in
escapeshellarg() function to sanitize the executed command’s arguments.
One last example is accepting foreign input to determine a file to load from the filesystem. This can be exploited by
changing the filename to a file path. You need to remove
"../", null bytes, or other characters from the
file path so it can’t load hidden, non-public, or sensitive files.
Sanitization removes (or escapes) illegal or unsafe characters from foreign input.
For example, you should sanitize foreign input before including the input in HTML or inserting it into a raw SQL query. When you use bound parameters with PDO, it will sanitize the input for you.
Sometimes it is required to allow some safe HTML tags in the input when including it in the HTML page. This is very hard to do and many avoid it by using other more restricted formatting like Markdown or BBCode, although whitelisting libraries like HTML Purifier exists for this reason.
It is dangerous to
unserialize() data from users or other untrusted sources. Doing so can allow malicious users to instantiate objects (with user-defined properties) whose destructors will be executed, even if the objects themselves aren’t used. You should therefore avoid unserializing untrusted data.
If you absolutely must unserialize data from untrusted sources, use PHP 7’s
allowed_classes option to restrict which object types are allowed to be unserialized.
Validation ensures that foreign input is what you expect. For example, you may want to validate an email address, a phone number, or age when processing a registration submission.
When creating configuration files for your applications, best practices recommend that one of the following methods be followed:
.phpextension. This ensures that, even if the script is accessed directly, it will not be output as plain text.
NOTE: As of PHP 5.4.0 the
register_globals setting has been removed and can no longer be used. This is only
included as a warning for anyone in the process of upgrading a legacy application.
When enabled, the
register_globals configuration setting that makes several types of variables (including ones from
$_REQUEST) available in the global scope of your application. This can easily lead to security
issues as your application cannot effectively tell where the data is coming from.
$_GET['foo'] would be available via
$foo, which can override variables that have not been declared.
If you are using PHP < 5.4.0 make sure that
register_globals is off.
Error logging can be useful in finding the problem spots in your application, but it can also expose information about the structure of your application to the outside world. To effectively protect your application from issues that could be caused by the output of these messages, you need to configure your server differently in development versus production (live).
To show every possible error during development, configure the following settings in your
Passing in the value
-1will show every possible error, even when new levels and constants are added in future PHP versions. The
E_ALLconstant also behaves this way as of PHP 5.4. - php.net
E_STRICT error level constant was introduced in 5.3.0 and is not part of
E_ALL, however it became part of
E_ALL in 5.4.0. What does this mean? In terms of reporting every possible error in version 5.3 it means you must
E_ALL | E_STRICT.
Reporting every possible error by PHP version
E_ALL | E_STRICT
To hide errors on your production environment, configure your
With these settings in production, errors will still be logged to the error logs for the web server, but will not be shown to the user. For more information on these settings, see the PHP manual:
Writing automated tests for your PHP code is considered a best practice and can lead to well-built applications. Automated tests are a great tool for making sure your application does not break when you are making changes or adding new functionality and should not be ignored.
There are several different types of testing tools (or frameworks) available for PHP, which use different approaches - all of which are trying to avoid manual testing and the need for large Quality Assurance teams, just to make sure recent changes didn’t break existing functionality.
Test-driven development (TDD) is a software development process that relies on the repetition of a very short development cycle: first the developer writes a failing automated test case that defines a desired improvement or new function, then produces code to pass that test and finally refactors the new code to acceptable standards. Kent Beck, who is credited with having developed or ‘rediscovered’ the technique, stated in 2003 that TDD encourages simple designs and inspires confidence.
There are several different types of testing that you can do for your application:
Unit Testing is a programming approach to ensure functions, classes and methods are working as expected, from the point you build them all the way through the development cycle. By checking values going in and out of various functions and methods, you can make sure the internal logic is working correctly. By using Dependency Injection and building “mock” classes and stubs you can verify that dependencies are correctly used for even better test coverage.
When you create a class or function you should create a unit test for each behavior it must have. At a very basic level
you should make sure it errors if you send it bad arguments and make sure it works if you send it valid arguments. This
will help ensure that when you make changes to this class or function later on in the development cycle that the old
functionality continues to work as expected. The only alternative to this would be
var_dump() in a test.php, which is
no way to build an application - large or small.
The other use for unit tests is contributing to open source. If you can write a test that shows broken functionality (i.e. fails), then fix it, and show the test passing, patches are much more likely to be accepted. If you run a project which accepts pull requests then you should suggest this as a requirement.
PHPUnit is the de-facto testing framework for writing unit tests for PHP applications, but there are several alternatives
Integration testing (sometimes called Integration and Testing, abbreviated “I&T”) is the phase in software testing in which individual software modules are combined and tested as a group. It occurs after unit testing and before validation testing. Integration testing takes as its input modules that have been unit tested, groups them in larger aggregates, applies tests defined in an integration test plan to those aggregates, and delivers as its output the integrated system ready for system testing.
Many of the same tools that can be used for unit testing can be used for integration testing as many of the same principles are used.
Sometimes also known as acceptance testing, functional testing consists of using tools to create automated tests that actually use your application instead of just verifying that individual units of code are behaving correctly and that individual units can speak to each other correctly. These tools typically work using real data and simulating actual users of the application.
There are two different types of Behavior-Driven Development (BDD): SpecBDD and StoryBDD. SpecBDD focuses on technical behavior of code, while StoryBDD focuses on business or feature behaviors or interactions. PHP has frameworks for both types of BDD.
With StoryBDD, you write human-readable stories that describe the behavior of your application. These stories can then be run as actual tests against your application. The framework used in PHP applications for StoryBDD is Behat, which is inspired by Ruby’s Cucumber project and implements the Gherkin DSL for describing feature behavior.
With SpecBDD, you write specifications that describe how your actual code should behave. Instead of testing a function or method, you are describing how that function or method should behave. PHP offers the PHPSpec framework for this purpose. This framework is inspired by the RSpec project for Ruby.
Besides individual testing and behavior driven frameworks, there are also a number of generic frameworks and helper libraries useful for any preferred approach taken.
PHP applications can be deployed and run on production web servers in a number of ways.
PaaS provides the system and network architecture necessary to run PHP applications on the web. This means little to no configuration for launching PHP applications or PHP frameworks.
Recently PaaS has become a popular method for deploying, hosting, and scaling PHP applications of all sizes. You can find a list of PHP PaaS “Platform as a Service” providers in our resources section.
If you are comfortable with systems administration, or are interested in learning it, virtual or dedicated servers give you complete control of your application’s production environment.
PHP, via PHP’s built-in FastCGI Process Manager (FPM), pairs really nicely with nginx, which is a lightweight, high-performance web server. It uses less memory than Apache and can better handle more concurrent requests. This is especially important on virtual servers that don’t have much memory to spare.
PHP and Apache have a long history together. Apache is wildly configurable and has many available modules to extend functionality. It is a popular choice for shared servers and an easy setup for PHP frameworks and open source apps like WordPress. Unfortunately, Apache uses more resources than nginx by default and cannot handle as many visitors at the same time.
Apache has several possible configurations for running PHP. The most common and easiest to setup is the prefork MPM with mod_php5. While it isn’t the most memory efficient, it is the simplest to get working and to use. This is probably the best choice if you don’t want to dig too deeply into the server administration aspects. Note that if you use mod_php5 you MUST use the prefork MPM.
Alternatively, if you want to squeeze more performance and stability out of Apache then you can take advantage of the same FPM system as nginx and run the worker MPM or event MPM with mod_fastcgi or mod_fcgid. This configuration will be significantly more memory efficient and much faster but it is more work to set up.
If you are running Apache 2.4 or later, you can use mod_proxy_fcgi to get great performance that is easy to setup.
If you find yourself doing manual database schema changes or running your tests manually before updating your files (manually), think twice! With every additional manual task needed to deploy a new version of your app, the chances for potentially fatal mistakes increase. Whether you’re dealing with a simple update, a comprehensive build process or even a continuous integration strategy, build automation is your friend.
Among the tasks you might want to automate are:
Deployment tools can be described as a collection of scripts that handle common tasks of software deployment. The deployment tool is not a part of your software, it acts on your software from ‘outside’.
There are many open source tools available to help you with build automation and deployment, some are written in PHP others aren’t. This shouldn’t hold you back from using them, if they’re better suited for the specific job. Here are a few examples:
Phing can control your packaging, deployment or testing process from within a XML build file. Phing (which is based on Apache Ant) provides a rich set of tasks usually needed to install or update a web application and can be extended with additional custom tasks, written in PHP. It’s a solid and robust tool and has been around for a long time, however the tool could be perceived as a bit old fashioned because of the way it deals with configuration (XML files).
Capistrano is a system for intermediate-to-advanced programmers to execute commands in a structured, repeatable way on one or more remote machines. It is pre-configured for deploying Ruby on Rails applications, however you can successfully deploy PHP systems with it. Successful use of Capistrano depends on a working knowledge of Ruby and Rake.
Ansistrano is a couple of Ansible roles to easily manage the deployment process (deploy and rollback) for scripting applications such as PHP, Python and Ruby. It’s an Ansible port for Capistrano. It’s been used by quite a lot of PHP companies already.
Rocketeer gets its inspiration and philosophy from the Laravel framework. Its goal is to be fast, elegant and easy to use with smart defaults. It features multiple servers, multiple stages, atomic deploys and deployment can be performed in parallel. Everything in the tool can be hot swapped or extended, and everything is written in PHP.
Deployer is a deployment tool written in PHP. It’s simple and functional. Features include running tasks in parallel, atomic deployment and keeping consistency between servers. Recipes of common tasks for Symfony, Laravel, Zend Framework and Yii are available. Younes Rafie’s article Easy Deployment of PHP Applications with Deployer is a great tutorial for deploying your application with the tool.
Magallanes is another tool written in PHP with simple configuration done in YAML files. It has support for multiple servers and environments, atomic deployment, and has some built in tasks that you can leverage for common tools and frameworks.
Managing and configuring servers can be a daunting task when faced with many servers. There are tools for dealing with this so you can automate your infrastructure to make sure you have the right servers and that they’re configured properly. They often integrate with the larger cloud hosting providers (Amazon Web Services, Heroku, DigitalOcean, etc) for managing instances, which makes scaling an application a lot easier.
Ansible is a tool that manages your infrastructure through YAML files. It’s simple to get started with and can manage complex and large scale applications. There is an API for managing cloud instances and it can manage them through a dynamic inventory using certain tools.
Puppet is a tool that has its own language and file types for managing servers and configurations. It can be used in a master/client setup or it can be used in a “master-less” mode. In the master/client mode the clients will poll the central master(s) for new configuration on set intervals and update itself if necessary. In the master-less mode you can push changes to your nodes.
Chef is a powerful Ruby based system integration framework that you can build your whole server environment or virtual boxes with. It integrates well with Amazon Web Services through their service called OpsWorks.
Continuous Integration is a software development practice where members of a team integrate their work frequently, usually each person integrates at least daily — leading to multiple integrations per day. Many teams find that this approach leads to significantly reduced integration problems and allows a team to develop cohesive software more rapidly.
– Martin Fowler
There are different ways to implement continuous integration for PHP. Travis CI has done a great job of making continuous integration a reality even for small projects. Travis CI is a hosted continuous integration service for the open source community. It is integrated with GitHub and offers first class support for many languages including PHP.
Running your application on different environments in development and production can lead to strange bugs popping up when you go live. It’s also tricky to keep different development environments up to date with the same version for all libraries used when working with a team of developers.
If you are developing on Windows and deploying to Linux (or anything non-Windows) or are developing in a team, you should consider using a virtual machine. This sounds tricky, but besides the widely known virtualization environments like VMware or VirtualBox, there are additional tools that may help you setting up a virtual environment in a few easy steps.
Vagrant helps you build your virtual boxes on top of the known virtual environments and will configure these environments based on a single configuration file. These boxes can be set up manually, or you can use “provisioning” software such as Puppet or Chef to do this for you. Provisioning the base box is a great way to ensure that multiple boxes are set up in an identical fashion and removes the need for you to maintain complicated “set up” command lists. You can also “destroy” your base box and recreate it without many manual steps, making it easy to create a “fresh” installation.
Vagrant creates folders for sharing your code between your host and your virtual machine, which means that you can create and edit your files on your host machine and then run the code inside your virtual machine.
If you need a little help to start using Vagrant there are some services that might be useful:
Docker - a lightweight alternative to a full virtual machine - is so called because it’s all about “containers”. A container is a building block which, in the simplest case, does one specific job, e.g. running a web server. An “image” is the package you use to build the container - Docker has a repository full of them.
A typical LAMP application might have three containers: a web server, a PHP-FPM process and MySQL. As with shared folders in Vagrant, you can leave your application files where they are and tell Docker where to find them.
You can generate containers from the command line (see example below) or, for ease of maintenance, build a
docker-compose.yml file for your project specifying which to create and how they communicate with one another.
Docker may help if you’re developing multiple websites and want the separation that comes from installing each on it’s own virtual machine, but don’t have the necessary disk space or the time to keep everything up to date. It’s efficient: the installation and downloads are quicker, you only need to store one copy of each image however often it’s used, containers need less RAM and share the same OS kernel, so you can have more servers running simultaneously, and it takes a matter of seconds to stop and start them, no need to wait for a full server boot.
After installing docker on your machine, you can start a web server with one command.
The following will download a fully functional Apache installation with the latest PHP version, map
/path/to/your/php/files to the document root, which you can view at
This will initialize and launch your container.
-d makes it runs in the background. To stop and start it, simply run
docker stop my-php-webserver and
docker start my-php-webserver (the other parameters are not needed again).
The command above shows a quick way to run a basic server. There’s much more you can do (and thousands of pre-built images in the Docker Hub). Take time to learn the terminology and read the Docker User Guide to get the most from it, and don’t run random code you’ve downloaded without checking it’s safe – unofficial images may not have the latest security patches. If in doubt, stick to the official repositiories.
The PHPDocker.io site will auto-generate all the files you need for a fully-featured LAMP/LEMP stack, including your choice of PHP version and extensions.
PHP is pretty quick by itself, but bottlenecks can arise when you make remote connections, load files, etc. Thankfully, there are various tools available to speed up certain parts of your application, or reduce the number of times these various time-consuming tasks need to run.
When a PHP file is executed, it must first be compiled into opcodes (machine language instructions for the CPU). If the source code is unchanged, the opcodes will be the same, so this compilation step becomes a waste of CPU resources.
An opcode cache prevents redundant compilation by storing opcodes in memory and reusing them on successive calls. It will typically check signature or modification time of the file first, in case there have been any changes.
It’s likely an opcode cache will make a significant speed improvement to your application. Since PHP 5.5 there is one built in - Zend OPcache. Depending on your PHP package/distribution, it’s usually turned on by default - check opcache.enable and the output of
phpinfo() to make sure. For earlier versions there’s a PECL extension.
Read more about opcode caches:
There are times when it can be beneficial to cache individual objects in your code, such as with data that is expensive to get or database calls where the result is unlikely to change. You can use object caching software to hold these pieces of data in memory for extremely fast access later on. If you save these items to a data store after you retrieve them, then pull them directly from the cache for following requests, you can gain a significant improvement in performance as well as reduce the load on your database servers.
Many of the popular bytecode caching solutions let you cache custom data as well, so there’s even more reason to take advantage of them. APCu, XCache, and WinCache all provide APIs to save data from your PHP code to their memory cache.
The most commonly used memory object caching systems are APCu and memcached. APCu is an excellent choice for object caching, it includes a simple API for adding your own data to its memory cache and is very easy to setup and use. The one real limitation of APCu is that it is tied to the server it’s installed on. Memcached on the other hand is installed as a separate service and can be accessed across the network, meaning that you can store objects in a hyper-fast data store in a central location and many different systems can pull from it.
Note that when running PHP as a (Fast-)CGI application inside your webserver, every PHP process will have its own cache, i.e. APCu data is not shared between your worker processes. In these cases, you might want to consider using memcached instead, as it’s not tied to the PHP processes.
In a networked configuration APCu will usually outperform memcached in terms of access speed, but memcached will be able to scale up faster and further. If you do not expect to have multiple servers running your application, or do not need the extra features that memcached offers then APCu is probably your best choice for object caching.
Example logic using APCu:
Note that prior to PHP 5.5, APC provides both an object cache and a bytecode cache. APCu is a project to bring APC’s object cache to PHP 5.5+, since PHP now has a built-in bytecode cache (OPcache).
Below is an example of how you might document a class with a few methods;
The documentation for the class as a whole has the @author tag and a @link tag. The @author tag is used to document the author of the code and can be repeated for documenting several authors. The @link tag is used to link to a website indicating a relationship between the website and the code.
Inside the class, the first method has a @param tag documenting the type, name and description of the parameter being passed to the method. Additionally it has the @return and @throws tags for documenting the return type, and any exceptions that could be thrown respectively.
The second and third methods are very similar and have a single @param tag as did the first method. The important
difference between the second and third methods’ doc block is the inclusion/exclusion of the @return tag.
@return void explicitly informs us that there is no return; historically omitting the
@return void statement also results in the same (no return) action.
It’s difficult to find interesting and knowledgeable PHP community members when you are first starting out. You can find an abbreviated list of PHP community members to get you started at:
To see which versions these PaaS hosts are running, head over to PHP Versions.
Rather than re-invent the wheel, many PHP developers use frameworks to build out web applications. Frameworks abstract away many of the low-level concerns and provide helpful, easy-to-use interfaces to complete common tasks.
You do not need to use a framework for every project. Sometimes plain PHP is the right way to go, but if you do need a framework then there are three main types available:
Micro-frameworks are essentially a wrapper to route a HTTP request to a callback, controller, method, etc as quickly as possible, and sometimes come with a few extra libraries to assist development such as basic database wrappers and the like. They are prominently used to build remote HTTP services.
Many frameworks add a considerable number of features on top of what is available in a micro-framework; these are called Full-Stack Frameworks. These often come bundled with ORMs, Authentication packages, etc.
Component-based frameworks are collections of specialized and single-purpose libraries. Disparate component-based frameworks can be used together to make a micro- or full-stack framework.
As mentioned above “Components” are another approach to the common goal of creating, distributing and implementing shared code. Various component repositories exist, the main two of which are:
Both of these repositories have command line tools associated with them to help the installation and upgrade processes, and have been explained in more detail in the Dependency Management section.
There are also component-based frameworks and component-vendors that offer no framework at all. These projects provide another source of packages which ideally have little to no dependencies on other packages, or specific frameworks.
For example, you can use the FuelPHP Validation package, without needing to use the FuelPHP framework itself.
Laravel’s Illuminate components will become better decoupled from the Laravel framework. For now, only the components best decoupled from the Laravel framework are listed above.
You can subscribe to weekly newsletters to keep yourself informed on new libraries, latest news, events and general announcements, as well as additional resources being published every now and then:
There are also Weeklies on other platforms you might be interested in; here’s a list of some.
There are many PHP books; sadly some are now quite old and no longer accurate. In particular, avoid books on “PHP 6”, a version that will now never exist. The next major release of PHP after 5.6 was “PHP 7”, partly because of this.
This section aims to be a living document for recommended books on PHP development in general. If you would like your book to be added, send a PR and it will be reviewed for relevancy.
The PHP community is as diverse as it is large, and its members are ready and willing to support new PHP programmers. Consider joining your local PHP user group (PUG) or attending larger PHP conferences to learn more about the best practices shown here. You can hang out on IRC in the #phpc channel on irc.freenode.com and follow the @phpc twitter account. Get out there, meet new developers, learn new topics, and above all, make new friends! Other community resources include the Google+ PHP Programmer community and StackOverflow.
If you live in a larger city, odds are there’s a PHP user group nearby. You can easily find your local PUG at
PHP.ug. Alternate sources might be Meetup.com or a search for
php user group near me
using your favorite search engine (i.e. Google). If you live in a smaller town, there may not be a
local PUG; if that’s the case, start one!
Special mention should be made of two global user groups: NomadPHP and PHPWomen. NomadPHP offers twice monthly online user group meetings with presentations by some of the top speakers in the PHP community. PHPWomen is a non-exclusive user group originally targeted towards the women in the PHP world. Membership is open to everyone who supports a more diverse community. PHPWomen provide a network for support, mentorship and education, and generally promote the creating of a “female friendly” and professional atmosphere.
The PHP community also hosts larger regional and national conferences in many countries around the world. Well-known members of the PHP community usually speak at these larger events, so it’s a great opportunity to learn directly from industry leaders.
ElePHPant is that beautiful mascot of the PHP project with elephant in their design. It was originally designed for the PHP project in 1998 by Vincent Pontier - spiritual father of thousands of elePHPants around the world and 10 years later adorable plush elephant toy came to birth as well. Now elePHPants are present at many PHP conferences and with many PHP developers at their computers for fun and inspiration.